CN102422622B - Signal mapping - Google Patents

Abstract

A circuit includes a signal processing circuit for accepting an input and for generating a set of outputs. The input is provided in an input range that has a set of representative values, and each output represents a measure of an association of the input with one or more of the representative values. The signal processing circuit includes a group of output sections, each output section being responsive to the input of the signal processing circuit. Each output section includes one or more sigmoid generators. Each sigmoid generator is responsive to an input of the output section to generate an output that represents a sigmoid function of the input of the output section. Each output section also includes a circuitry for combining the outputs of the one or more sigmoid generators to form one of the set of outputs of the signal processing circuit. An input transformation circuit is coupled to the plurality of output sections. The input transformation circuit is configurable to transform the input of the signal processing circuit for controlling a mapping characteristic from the input to the set of outputs.

Description

Signal image technology

The mutual reference of related application

The name that this application opinion is filed an application on March 2nd, 2009 is called the U.S.Provisional Serial the 61/156th of " Signal Mapping ", No. 721 and the name of filing an application on January 8th, 2010 are called the U.S.Provisional Serial the 61/293rd of " Signal Mapping ", the interests of No. 439, the content of the two is incorporated to thinks reference data herein.

The name that this application also relates on March 2nd, 2009 files an application is called the U.S.Provisional Serial the 61/156th of " Belief Propagation Processor ", the U.S.Provisional Serial the 61/156th that the name that on March 2nd, No. 721 1 files an application is called " Circuits for Soft Logical Functions ", No. 794 and the name of filing an application on March 2nd, 2009 are called the U.S.Provisional Serial the 61/156th of " Circuits for Soft Logical Functions ", No. 735.The content of these applications above-mentioned is incorporated to thinks reference data herein.

Technical field

The present invention relates to signal image technology.

Background technology

This specification relates to signal image technology, such as, relate to the acceptor circuit for communication system.

Communication system can be used for the information exchange between user.These systems allow extensive apart from interior and transmit data via dissimilar medium (such as, air and line).Various coding techniques can in order to improve communication performance.Similar coding techniques can be used for stocking system.

Fig. 1 shows an example of communication system 100, and it comprises reflector 110, communication channel 120, and receiver 130.Reflector 110 receives input (such as, digital bit string) and is modulated to applicable form (such as, such as electromagnetic analog signal) to transmit in channel 120 (such as, air).Receiver 130 receives the signal of this transmission and is demodulated into original digital form to export.

In some applications, signal is encoded before transmission for various purposes.The error that this application relates to for the signal sent when there is interchannel noise reduces and/or corrects.In these cases, encoder 112 can in order to be incorporated into the redundancy through design in this input signal to produce coding bit string (being sometimes also called code word) carefully.Projection instrument 114 is set up one and is directly videoed between this coding bit string and a group code, and wherein each symbol corresponds to the analog value of one (or multiple) entity attribute of this analog signal that need transmit in this channel.At this receiver end, once this transmission signal is received, de-mapping device 134 just extracts position estimated value from measured value, and a decoder 136 (or in some other examples, the soft iterative equaliser for such as viterbi equalizer) subsequently by this estimated value decoding to recover this digital signal.

Because this signal received may by the impact of interchannel noise, so in some examples, it may be useful for first noise signal being interpreted as " soft " position and then these soft positions being decoded as numerical data.Soft position provides the reliability of contraposition estimated value-with regard to binary coding, is measurement mode institute's received bit to the probability of the value being zero or.

Summary of the invention

General aspects more of the present invention relate to such as, are configured to the circuit of information extraction from noise inputs.This circuit comprise in order to receive input and in order to produce one group export signal processing circuit.This input is provided in the input range with a class value, and each exports the measuring value associated representing this input and one or more typical value of this typical value.

This signal processing circuit comprises one group of output, and each output is in response to this input of this signal processing circuit.Each output comprises one or more tangent bend function generator of more vairable.Each tangent bend function generator of more vairable is in response to the input of this output to produce output, and this output represents the tangent bend function of this input of this output.Each output also comprises circuit, and it is in order to combine this output of this one or more tangent bend function generator of more vairable to the output in this group output forming this signal processing circuit.

Input change-over circuit is coupled to the plurality of output.This input configurablely in order to change this signal processing circuit of this input change-over circuit control from this input to this group export a map feature.

Embodiment can comprise one or more feature in following characteristics.

This output of this tangent bend function generator of more vairable can represent the second tangent bend function of this input arriving this signal processing circuit.

This map feature inputing to the output of this group from this can be associated with the characteristic of this second tangent bend function.

This characteristic of this second tangent bend function can comprise the position of the transition point of this tangent bend function and/or the rate of change at this transition point place of this tangent bend function.

In some examples, at least one output comprises two tangent bend function generators of more vairable.

Each output also comprises an output conversion circuit, and it controls to input to from this second map feature that this group exports in order to this output of one or more tangent bend function generator of more vairable corresponding to photograph configurablely.

Each tangent bend function generator of more vairable can use differential pair circuit to configure.This differential pair circuit can comprise pair of transistor, and each transistor is configured this input receiving this tangent bend function generator of more vairable with the corresponding polarity in bipolarity.

This differential pair circuit can comprise a pair mosfet transistor being configured to operate under overcritical Value Operations state.Selectively, this differential pair circuit can comprise a pair bipolar junction transistor.

This input arriving this signal processing circuit can comprise and representing in communication channel with the analog signal of the numeric string of sign format transmission.This analog signal can comprise the noise component(s) determined by this channel at least partly.Such as, this analog signal can comprise Gaussian noise component.

Each typical value of this input range can be associated to the corresponding symbol in discrete one group of possible symbol.

In some examples, each of this signal processing circuit exports the corresponding probability that this transmission symbol of expression corresponds to the symbol in this possibility symbol.In some other examples, each of this signal processing circuit exports and represents that one of the numeric string corresponding numeral transmitted has the corresponding probability of the probable value in one group of discrete probable value.The numeric string transmitted can be binary string.In some other examples, this output of this signal processing circuit represents the respective probability of the corresponding numerical digit of binary number, the image pixel intensities level before this binary number representation noise adds in image.In some other examples, any variable paid close attention to can be used as this image pixel intensities used in an in the end example and uses.

In some examples, each output of this signal processing circuit can comprise transmitted position or the expression of symbol in log-domain.In some examples, each output of this signal processing circuit can comprise this or the expression of symbol in log-domain of the state of this variable that expression is paid close attention to.

This input arriving this signal processing circuit can the form of differential signal be implemented.Each input arriving this output can also the form of differential signal be implemented.

In another general aspect, signal processing circuit be provided to receive input and in order to produce one group of output.This input is in the input range comprising a class value.Each exports the measuring value associated representing this input and one or more typical value in this typical value.

This signal processing circuit comprises multiple part, every part produce this group export in a corresponding output.Every part comprises one or more input change-over circuit, and this one or more input change-over circuit this input configurablely in order to change this signal processing circuit controls to input to the map feature of this corresponding output from this.Every part also comprises one or more tangent bend function generator of more vairable, and each tangent bend function generator of more vairable is coupled to corresponding input change-over circuit and is configured to receive the input after conversion from this corresponding input change-over circuit and produce the tangent bend function output of the tangent bend function representing this input.Every part also comprise in order to combination from this tangent bend function generator of more vairable this tangent bend function export formed this signal processing circuit this group export in one export circuit.

Embodiment can comprise one or more feature in following characteristics.

Each output during this group exports comprises the expression of this input in log-domain, the log-likelihood ratio of such as this input.

Every part also comprises one or more output conversion circuit.Each output conversion circuit is coupled to the corresponding tangent bend function generator of more vairable and exports in order to this tangent bend function corresponding to photograph configurablely and controls to input to the second map feature of this corresponding output from this.

In another general aspect, provide in order to receive input and to produce one group of method exported.This input is in the input range comprising a class value.Each exports the measuring value associated representing this input and one or more typical value in this typical value.The method comprise use input change-over circuit to change this input with control from this input to this group export map feature.

The method also comprises, and in each output of multiple output, at least one the tangent bend function producing the tangent bend function of the input after representing conversion exports, and each tangent bend function exports and produced by the corresponding tangent bend function generator of more vairable.The method also comprises, and in each output of the plurality of output, uses combinational circuit to combine this tangent bend function and exports with the output formed in the output of this group.

Embodiment can comprise one or more feature in following characteristics.This output of this tangent bend function generator of more vairable represents the second tangent bend function of this input.Input to this map feature that this group exports be associated from this with the characteristic of this second tangent bend function, the rate of change at the position of a transition point of such as this tangent bend function and the transition point place of this tangent bend function at least one of them.

The method also comprises, in each output of the plurality of output, use corresponding output conversion circuit to change each tangent bend function export control from this input to this group export the second map feature.

Embodiment can comprise one or more advantage in following advantages.

In certain embodiments, provide to use board design transmission signal to be demodulated into the system and method for soft position.This board design does not need input digitlization and can produce " soft " position (such as, position probability, or the rpm value of position estimated value and these estimations) based on this input value observed by this receiver.These soft positions, once produce in an analogue form, can be supplied to follow-up decoder (such as, error corrector) by the decoding of transmission digit order number.

In certain embodiments, this board design uses two stage treatment to produce soft position: symbol detection and position probability calculate.Symbol detection is implemented by using analog circuit (the tangent bend function generator of more vairable such as formed based on differential pair circuit) to be compatibly similar to the amount of analysis derived for symbol likelihood.This analog circuit can adopt various types of transistor, such as bipolar junction transistor (BJT), the mos field effect transistor (MOSFET) operated in weak inversion state (that is, sub-threshold pattern) and the MOSFET operating in strong inverted status (i.e. overcritical binarization mode).These analog circuits operationally can use various types of transistor in the circuit, and not necessarily need the transistor presenting one Exponential current-voltage relationship.In some examples when using overcritical value MOSFET, circuit speed can improve greatly.

In certain embodiments, this system and method in order to transmission signal is demodulated into symbol likelihood, and can not necessarily produce soft position.

In some examples, each symbol likelihood can calculate via a part for the one group of nextport hardware component NextPort be separated completely instead of a nextport hardware component NextPort.This modularization execution mode allows this board design can configure for various should being used for, and such as needs the application of the quantization level of the different numbers that the even of quantization level or non-uniform spacing need be utilized to detect.

In numerous applications, this noise signal is interpreted as " soft " symbol and then these soft symbols are balanced or decoding or to be otherwise treated to data be useful.Except a communication system, such as also can by noise image data de-mapping (demap).A pixel in image can be considered one 8 fixing point variablees of a strength level that can present in 256 varying strength levels.In some cases, we may know Gauss's (or other kinds) noise joined in this image due to the thermal noise in order to catch in the CMOS of this image.When knowing variance or other characteristics of this noise profile, the primary signal from this image sensor can be converted to the probability of each in these 256 possible strength levels by a demapper.

Other features of the present invention and advantage as can be seen from following explanation and accessory rights require.

Embodiment

1 system survey

Following explanation provides the example of signal processing circuit, and such as it can use in for the receiver of communication system.

Communication system can use signal madulation (such as, the being expressed as bit string) information on abstract sense to be converted to the entity attribute of transmission signal.Direct relation can be set up between bit string and symbol, and wherein each symbol corresponds to a typical value of this entity attribute of this transmission signal.Such as, when data transmit as electromagnetic wave in atmosphere, each symbol can represent a level of one or more string and the physical parameter (such as, amplitude, phase place and frequency) corresponding to ripple.

Various signal modulation technique can use in communication system.Example comprises pulse-amplitude modulation (PAM) and quadrature amplitude modulation (QAM).The unrestriced object for explanation, following example is mainly described the background of modulating based on the PAM of the reflection (symbol stellar map) of Gray code of set using the coded representation from character string to this symbol.Other modulation techniques comprise binary phase shift keying (BPSK) and frequency shift keying (FSK).

Fig. 2 display is configured to an embodiment of the acceptor circuit 200 used together with this communication system of Fig. 1.Acceptor circuit 200 comprises demapper 210, and expression one is worth a noise input signal of ν by it, and the expression one such as received from receiver antenna transmits a voltage code signal of bit string, is converted to " soft " position.In some examples, a decoder 220 then uses this soft position to make " rigid " decision-making, such as, in order to produce the digit order number representing this transmission bit string.

In some examples, in order to obtain this soft position, demapper 210 implements secondary simulation process.More specifically, a symbol likelihood generator 212 produces one group of signal likelihood P (S according to this input ν of the displacement corresponding to reflector place _k), wherein each P (S _k) represent that this is input as a given symbol S _kprobability.Subsequently, a probability generator 214 uses these symbol likelihoods produced to produce soft position P (b _i), wherein each P (b _i) represent that i-th numeral of this bit string transmitted is the probability (assuming that using binary code in this example) of zero or.

Fig. 2 is also presented at an example of a PAM4 (the i.e. 4 rank Modulation and Amplitude Modulation) stellar map forming a reflection between bit string and symbol.Herein, each two numerical digit string (b ₀, b ₁) corresponding to four symbol S ₀to S ₃in a symbol, such as, character string (0,0) is corresponding to symbol S ₀.Each symbol is associated designed to a typical value of this input, such as, and S ₀be associated with value-1.

Ideally, during data transmit, there is not any noise source or distortion sources, when observing input ν=-1/3, knowing transmission symbol is S ₁and therefore correspond to transmission bit string (0,1).But in actual applications, the value of this input ν is subject to the impact of interchannel noise and/or other source of signal distortion usually.Therefore, when a given observation input value ν, be probabilistic to the judgement of this transmission symbol in nature.

More specifically, the model M of this reflector given and this channel, and given reception value ν, then de-mapping device 210 is determined to correspond to one to the probability of the input of location vector b:

p(b|ν，M)＝p(b ₀，b ₁|ν，M)

Use bayesian rule, aforesaid equation can Factorization be

p(b ₀，b ₁，ν|M)＝p(b ₀|M)p(b ₁|M)p(ν|b ₀，b ₁，M)。

Factor p (b ₀| M) and p (b ₁| M) represent b ₀and b ₁priori probability and the regular message of block (such as, decoder 220) after equaling from the demapper 210 in signal processing path.If priori probability is unavailable, then b ₀and b ₁priori probability be assumed to and be uniformly distributed.

By the equational mathematical computations of above-mentioned Factorization, an equation group can be determined, this probability unique definition is the function of input data, noise parameter and the priori information available about place value by it.In this manual, an approximate execution mode of these relations using the electronic circuit being based on transistor is described.

Fig. 3 A shows the theoretical prediction (such as, probability a posteriori) to the symbol likelihood of the function as this input value, assuming that this channel is by an additive white Gaussian noise (AWGN; With one standard deviation=0.3 illustrate) force at these transmission data and each symbol through transmit possibility equal.Especially, this AWGN can be expressed as

$P\left(x\right)=\frac{1}{\mathrm{\σ}\sqrt{2\mathrm{\π}}}\exp (-\frac{{(x-\mathrm{\μ})}^2}{2{\mathrm{\σ}}^2}),$

Wherein σ is standard deviation and average μ is presumed to be zero.

In this figure, have 4 overlapping districts, each overlapping district is by a respective symbol grayscale curve P (S _k) define.Each P (S _k) to be shown in this input measured value be that under the condition of ν, this transmission signal is S to curve table _kprobability.This symbol grayscale curve is determined from the ratio of the average sum of the gauss of distribution function of ν centered by expected character position and the gauss of distribution function of this ν and each character position.In general, this symbol grayscale curve is divided into two classifications.First category comprises and represents two ends symbol (S ₀and S ₃) dullness (such as, S shape) the symbol likelihood function of probability.Second classification comprises and represents intermediate symbols (S ₁and S ₂) mode (such as convex shaped, although not necessarily symmetrical) the symbol likelihood function of probability.Each in this mode curve represents input value place at this of (or close) corresponding this symbol to be had a peak value and exports.Such as, by curve P (S ₁) close the region (shadow region as in this figure) irised out there is the mode be positioned near input value-1/3, indicate when a given observation input-1/3, this transmission symbol has the chance of about 85% to be symbol S ₁.Position b ₀and b ₁probability determine by marginalizing within the scope of the likelihood of derivation symbol.

It is to be noted, in this example, this probability curve is regular.That is, under any input value, all 4 likelihood P (S ₀) to P (S ₃) and equal 1.This is not necessarily correct all the time in other examples.

When data transmit in the communication channel of different qualities, these grayscale curves P (S _k) center substantially maintain near this class value of defining in this stellar map.But the shape of each curve, may change according to noise characteristic.

Fig. 3 B, such as, shows another prediction to the symbol likelihood as input function when this channel has a Gaussian noise of variance less (σ=0.1).Herein, each P (S _k) curve presents a nearly rectangular shape, it has than previously shown in figure 3 a this corresponding P (S _k) side slope that curve is steeper.Overlapping between zones of different also reduces in the case greatly.In other words, a channel comparatively cleaned allows for a given input, more reliably determines better symbol.It is to be noted that another is extremely, as σ trend infinite (namely noise is uniformly distributed), all symbol probabilities are equal, and probability is 1/M (situation for there being M symbol in this corresponding symbol stellar map).

2 symbol likelihood generators

Following part provides the example of a symbol likelihood generator, and it uses analog circuit to be configured to the symbol likelihood function of this theoretical prediction producing approximate diagram 3A and 3B.As discussed previously, there is the symbol likelihood function of two kinds, mode (i.e. convex shaped) function of dullness (the i.e. S shape) function namely representing the probability of two ends symbol and the probability representing intermediate symbols.The configurable tangent bend function generator of more vairable such as can be used to build, described in detail by hereafter in order to provide the circuit of the transfer function of this two type.

In this description, term " the tangent bend function generator of more vairable (sigmoid generator) " refers to that its transfer function is formed or is similar to circuit or the circuit unit of a pair of function of flexure.In general, a pair of function of flexure is real number value type and is differentiable, and it has a non-negative or an anon-normal first derivative and a lucky flex point (namely the second dervative of this function changes this point of sign whereabouts).The a pair of function of flexure has two asymptotes, namely when input is just tending to and bear infinite.The a pair of function of flexure can adopt various forms, comprises, such as, and logarithmic function, ordinary arc tangent, tanh and error function.

2.1 circuit that MODAL TRANSFORMATION OF A function is provided

Fig. 4 shows the calcspar of a circuit 400, and it provides this symbol likelihood function P (S of approximate diagram 3A and 3B ₁) to P (S ₂) a MODAL TRANSFORMATION OF A function.Herein, circuit 400 comprises the tangent bend function T being configured to generation respectively and can being controlled by external parameter ₁(ν) and T ₂(ν) a pair configurable tangent bend function generator of more vairable 410 and 410 '.

Such as, the first configurable tangent bend function generator of more vairable 410 comprises providing a core tangent bend function generator of more vairable 416 of a pair of function of flexure S (ν) (its flex point is positioned at initial point place), in order to make this tangent bend function S (ν) along trunnion axis displacement one distance a ₁an adder 412, and in order to by this tangent bend function a slope change into b ₁a multiplier 414 doubly.In other words, the output transfer function T of block 410 ₁the a that has been shifted ₁and through extending to b ₁doubly/boil down to 1/b ₁one amendment tangent bend function, make T ₁(ν)=S ((a ₁+ ν) b ₁).

Similarly, the second configurable tangent bend function generator of more vairable 410 ' produces a that has been shifted ₂and through extending to b ₂doubly/boil down to 1/b ₂one amendment tangent bend function T ₂, just this output tangent bend function is also anti-phase by an inverter 418.

By the output of this first and second configurable tangent bend function 410 and 410 ' being combined in an adder 430, circuit 400 produces this mode P (S of total transfer function T, its approximate diagram 3A ₁) and P (S ₂), wherein T (ν)=S ₁((a ₁+ ν) b ₁)-S ₂((a ₂+ ν) b ₂).Herein, the characteristic of this T curve is by adjusting this four parameter a ₁, b ₁, a ₂and b ₂in the value of one or more parameter control, such as, consider the impact of various interchannel noise on this transmission signal.In some examples, this two configurable tangent bend function generator of more vairable 410 and 410 ' is configured at identical output level place saturated for the object adjusted.

Fig. 5 A show needle is to a circuit structure 500 of an execution mode of circuit 400 shown in Fig. 4.In this example, circuit 500 comprises a folders 520 of use a pair tangent bend function generator of more vairable 516 and 516 ' (block 416 and 416 ' corresponding to Fig. 4), and it is such as by using differential pair circuit to be formed.Circuit 500 also comprises side-play amount a ₁and a ₂be incorporated into a pair amplifier in this transfer function T of Fig. 4 512 and 512 ' and a pair variable gain amplifier (VGA) 514 and 514 ' in order to the slope that adjusts this transfer function T.Amplifier 512 and 512 ', each used analog circuit in VGA 514 and 514 ' and folders 520 implements.In other words, the input of circuit 500 and output are continuous print variable signal (such as, in voltage or the form of current signal).

Fig. 5 B shows an example of the folders 520 of use one folding amplifier 520A.Folding amplifier 520A comprises two differential pair circuit, and its Differential Input offsets relative to each other and exports difference current and reduced.Each differential pair circuit comprises three basic module: pair of transistor M ₁and M ₂, a pair loading resistor R, and make electric current (such as, 1/2I _bias) can be used for the current source switching this transistor.This transfer function of differential pair circuit has a pair of curve form and thus can by various analysis and specific function approximate representation.

For purposes of illustration, transistor M ₁and M ₂the transistor of other types (such as bipolar junction transistor) adopts the form of mosfet transistor in this figure, although also can be used.This MOSFET can operate under weak inversion state (i.e. sub-threshold pattern), or selectively, can operate under strong inverted status (i.e. overcritical binarization mode).This load R can implement in every way, the mosfet transistor comprise discrete resistor, operating under a linear condition, or an additional current source.

In this example, folding amplifier 520A uses fully differential signaling, and namely (such as, each differential pair circuit receives the input of differential voltage form ) and output (such as, the OUT of generation differential voltage/current forms ⁺-OUT ^-).Differential signal is associated with a common mode, and the input port for each differential pair circuit is defined as one similar definitions is applicable to output port.Usually, common mode electrical level remains constant in whole signal chains.Selectively, this circuit can repel (namely to reduce the sensitivity of circuit to common mode variations) through design to provide common mode.Such as, differential pair can be set in I _biasone fixed common mode of R/2 uses together.

Figure 22 A shows the calcspar of one second embodiment of a circuit 2200, and it provides this symbol likelihood function P (S of approximate diagram 3A and 3B ₁) and P (S ₂) a MODAL TRANSFORMATION OF A function.Herein, circuit 2200 comprises the tangent bend function T being configured to generation respectively and can being controlled by external parameter ₁(ν) and T ₂(ν) a pair configurable tangent bend function generator of more vairable 2210 and 2210 '.

Such as, the first configurable tangent bend function generator of more vairable 2210 comprises providing a core tangent bend function generator of more vairable 2216 of a pair of function of flexure S (ν) (its flex point is at initial point place), in order to make this tangent bend function S (ν) along trunnion axis displacement one distance a ₁an adder 2212, and in order to by this tangent bend function a slope change into b ₁a multiplier 2214 doubly.In other words, the output transfer function T of block 2210 ₁the a that has been shifted ₁and through extending to b ₁doubly/boil down to 1/b ₁one amendment tangent bend function, make T ₁(ν)=b ₁s (a ₁+ ν).

Similarly, the second configurable tangent bend function generator of more vairable 2210 ' produces a that has been shifted ₂and through extending to b ₂doubly/boil down to 1/b ₂one amendment tangent bend function T ₂, just this output tangent bend function is also anti-phase by an inverter 2218.

By being combined in an adder 2230 by this first and second configurable tangent bend function generator of more vairable 2210 and 2210 ', circuit 2200 produces this mode P (S of total transfer function T, its approximate diagram 3A ₁) and P (S ₂), wherein T (ν)=b ₁s ₁(a ₁+ ν)-b ₂s ₂(a ₂+ ν).Herein, the characteristic of this T curve, by such as, adjusts this four parameter a ₁, b ₁, a ₂and b ₂in the value of one or more parameter control, to consider the impact of various interchannel noise on this transmission signal.In some examples, this two configurable tangent bend function generator of more vairable 2210 and 2210 ' is configured at identical output level place saturated for the object adjusted.

Figure 22 B show needle is to a circuit structure 2250 of an execution mode of circuit 2200 shown in Figure 22 A.In this example, circuit 2250 comprises a folders 2260 of use a pair tangent bend function generator of more vairable 2256 and 2256 ' (block 2216 and 2216 ' corresponding to Figure 22 A), and it is such as by using differential pair circuit to be formed.Circuit 2250 also comprises side-play amount a ₁and a ₂be incorporated into a pair adder in this transfer function T of Figure 22 A 2252 and 2252 ', and in order to a pair variable gain amplifier (VGA) 2254 and 2254 ' of this slope of adjusting this transfer function T.Adder 2252 and 2252 ', each used analog circuit in VGA 2254 and 2254 ' and folders 2260 implements, the input of circuit 2250 and output is made to be continuous print variable signal (such as, in voltage or the form of current signal).In some examples, folders 2260 is implemented as a folding amplifier of folding amplifier 520A shown in such as Fig. 5 B.

When not being restricted, for conforming object, example is herein described as using full-differential circuits.Note that the folding amplifier using single ended input is also possible, as long as the form of input and output signal is consistent with the operation of residual circuit assembly.In addition, difference signaling may be favourable because its dynamic range be greater than single-ended signaling and noise immunity higher than single-ended signaling.In some examples, use electric current may be more convenient as output signal than use voltage as output signal.In these cases, the afterbody produced in the signal transacting of output voltage is responsible for setting up correct output common mode level.

Herein, the difference output (current encoded or voltage code signal) of folders 520 is configured to represent symbol likelihood P (S _k) (such as, k=1 or 2).In some examples, the amplitude of this difference output and P (S _k) become direct direct ratio.In some other examples, utilize the alternate embodiments of folders 520, the amplitude of this difference output can with the P (S with a DC side-play amount _k) linear.

2.2 circuit that monotonic transitions function is provided

Fig. 6 shows the calcspar of a circuit 600, and it provides can in order to the symbol likelihood function P (S of approximate diagram 3A and 3B ₀) and P (S ₃) a S shape transfer function.Herein, circuit 600 comprises in order to produce the tangent bend function T that can be controlled by external parameter ₀a configurable tangent bend function generator of more vairable 610 and in order to along vertical axis by this tangent bend function T ₀displacement one distance K ₀a reference block 640.The tangent bend function generator of more vairable 610 comprises a side-play amount a ₀be incorporated into the amplifier 612 in a pair of function of flexure produced by a generator 616, in order to the slope of tangent bend function 616 is adjusted to b ₀a variable gain amplifier 614 doubly and in order to make this function reverse or do multiplying to produce tangent bend function T ₀a circuit unit 618.In some cases, circuit unit 618 is inverters; In other cases, assembly 618 is Configurable Multipliers, and this function is multiplied by+1 or-1 by it.By the output of block 610 and 640 being combined in an adder 630, circuit 600 produces this S shape P (S of total transfer function T, its approximate diagram 3A and 3B ₀) and P (S ₃).

Fig. 7 A show needle is to a circuit structure 700 of an execution mode of circuit 600 shown in Fig. 6.In this example, circuit 700 comprises the folders 720 using the tangent bend function generator of more vairable 716 (block 616 corresponding to Fig. 6), and it is such as by using differential pair circuit to be formed.One constant reference 740 (corresponding to block 640) is coupled to the tangent bend function generator of more vairable 716 and exports displacement to make this tangent bend function.Circuit 700 also comprises introducing Input Offset Value a ₀an amplifier 712 and a VGA 714 of this slope in order to this transfer function T of adjusting Fig. 6.Equally, each the used analog circuit in amplifier 712, VGA 714 and folders 720 is implemented.In other words, the input of circuit 700 and output are continuous print variable signal.

Fig. 7 B shows an example of the folders 720 of use one differential pair circuit 720A.Differential pair circuit 720A is coupled to a current source 740A, and it makes this output signal offset a fixed amount as constant reference 740.Herein, by current source 740A being set in level (the i.e. 1/2I identical with this current source in differential pair circuit 720A _bias), the output of this folders can upward displacement to being all the shape maintaining this tangent bend function just simultaneously.Note that if desired, this direction of displacement is by reversing to replace fixed current source 740A with the fixed current place being coupled to cathode output end.Equally, this output of folders 720 represents symbol likelihood P (S _k) (such as, k=0 or 3), direct direct ratio or selectively linear with a DC side-play amount is become with a DC side-play amount.

There is provided another example of the folders 720 of displacement tangent bend function that the folding amplifier 520 of Fig. 5 B can be used by one group of differential input signal is set to constant to configure.In other words, the circuit in this differential pair circuit is effectively as the current source being coupled to another differential pair circuit.

One the 3rd example (not shown) of folders 720 is implemented one and is substituted differential pair circuit, and himself produces and this symbol likelihood P (S _k) a proportional non-negative tangent bend function exports.

Figure 23 A shows the calcspar of another embodiment of a circuit 2300, and it provides can in order to this symbol likelihood function P (S of approximate diagram 3A and 3B ₀) and P (S ₃) a S shape transfer function.Herein, circuit 2300 comprises in order to produce the tangent bend function T that can be controlled by external parameter ₀a configurable tangent bend function generator of more vairable 2310 and in order to along vertical axis by this tangent bend function T ₀displacement one distance K ₀a reference block 2340.The tangent bend function generator of more vairable 2310 comprises a side-play amount a ₀be incorporated into the amplifier 2312 in the tangent bend function produced by generator 2316, in order to the slope of tangent bend function 2316 is adjusted to b ₀variable gain amplifier 2314 doubly and in order to make this function reverse or do multiplying to produce tangent bend function T ₀circuit unit 2318.In some cases, circuit unit 2318 is inverters; In other cases, assembly 2318 is Configurable Multipliers, and this function is multiplied by+1 or-1 by it.By the output of block 2310 and 2340 being combined in adder 2330, circuit 2300 produces total transfer function T, and it is similar to the S shape P (S of Fig. 3 A and 3B ₀) and P (S ₃).

Figure 23 B show needle is to a circuit structure 2350 of an execution mode of circuit 2300 shown in Figure 23 A.In this example, circuit 2350 comprises the folders 2360 using the tangent bend function generator of more vairable 2356 (block 2316 corresponding to Figure 23 A), and it is such as by using differential pair circuit to be formed.Constant reference 2370 (corresponding to block 2340) is coupled to the tangent bend function generator of more vairable 2356 and exports displacement to make this tangent bend function.Circuit 2350 also comprises introducing Input Offset Value a ₀amplifier 2352 and a VGA 2354 of this slope in order to this transfer function T of adjusting Figure 23 A.Equally, each the used analog circuit in amplifier 2352, VGA 2354 and folders 2360 is implemented.In other words, the input of circuit 2350 and output are continuous print variable signal.In some cases, folders 2360 utilizes a differential pair circuit of differential pair circuit 720A shown in such as the 7th figure to implement.

2.3 for the demonstrative circuit of symbol likelihood generator

Fig. 8 show needle is to a demonstrative circuit structure 800 of a symbol likelihood generator of use this circuit engineering above-mentioned.This symbol likelihood generator receives a Differential Input ν ⁺, ν ^-and produce one group of output P (S ₀) to P (S ₃), this output each represents the probability that this input symbol corresponding with is associated.

More specifically, this symbol likelihood generator 800 comprises one group of folders 820A-D, and each folders is in response to this input of this symbol likelihood generator.Each folders comprises one or two tangent bend function generators of more vairable (being shown as SG).Each SG produces in response to the input of this corresponding folders the output representing this tangent bend function S inputted.The output that each folders also comprises combining its SG exports P (S to form these _k) the circuit (such as, adder) of corresponding output.

Symbol likelihood generator 800 also comprises input change-over circuit 810, and it is coupled to this group folders in order to conversion input ν ⁺, ν ^-.Input change-over circuit 810 comprises one group of amplifier 812A-812C and can in order to adjust the output of symbol likelihood generator 800 to emulate the one group VGA 814A-814C of symbol likelihood function to the parameter dependence (such as, to the dependence of σ) of interchannel noise.

Figure 24 show needle is to an alternate embodiment of a circuit structure 2400 of a symbol likelihood generator of use this circuit engineering above-mentioned.This symbol likelihood generator receives a Differential Input ν ⁺, ν ^-and produce one group of output P (S ₀) to P (S ₃), this output each represents the probability that this input symbol corresponding with is associated.

More specifically, symbol likelihood generator 2400 comprises one group of folders 2420A-D, and each folders is in response to this input of this symbol likelihood generator.Each folders comprises one or two tangent bend function generators of more vairable (being shown as SG).Each SG produces in response to the input of this corresponding folders the output representing this pair of function of flexure S inputted.The output that each folders also comprises combining its SG exports P (S to form these _k) the circuit (such as, an adder) of a corresponding output.

Symbol likelihood generator 2400 also comprise be coupled to this group folders in order to conversion input ν ⁺, ν ^-one group of amplifier 2212A-2212C.In addition, one group of VGA 2414A-2414C is in order to adjust the output of symbol likelihood generator 2400 to emulate the parameter dependence (such as, to the dependence of σ) of symbol likelihood function to interchannel noise.

3 probability generators

Once this symbol likelihood generator P (S ₀) to P (S ₃) produce for a specific input, position probability can also calculate by an analog form, such as, and the mode of being sued for peace by signal.

Fig. 9 shows the calcspar of an example of the probability generator 900 used together with this symbol likelihood generator of Fig. 8.Based on previously shown in fig. 2 Gray code stellar map, this probability can obtain according to symbol likelihood as follows:

P(b ₀＝0)＝P(S ₀)+P(S ₁) (1a)

P(b ₀＝1)＝P(S ₂)+P(S ₃) (1b)

P(b ₁＝0)＝P(S ₀)+P(S ₃) (1c)

P(b ₁＝1)＝P(S ₁)+P(S ₂) (1d)

Each in this four equation can perform in a corresponding adder in adder 910,912,914 and 916.

It is to be noted, due to inherent limitations condition, be not (1a)-all four calculating of (1b) need perform to obtain a probability.Such as, with regard to binary coding, it is known that position b _ibe 1 probability and be 0 probability to add up be 100%, this means once obtain P (b _i=0), also known P (b _i=1).Therefore, position probability generator 900 can be reduced to produce P (b ₀=0) and P (b ₁=0) one group of two adder (such as, 910 and 912).

In analog circuit execution mode, input and the output of position probability generator 900 can be rendered as voltage, or are selectively rendered as electric current.In some applications, current signal is used may to be useful because the summation of current forms uses Kirchhoff (Kirchhoff) current law easily to realize.In these cases because a symbol likelihood can calculate not coordination probability time used more than once, therefore represent that multiple copies of the same current of symbol likelihood may be expected.There is some modes duplicate current signal in board design.The same units gain buffer that one example is driven by identical signal for use.Another example for implement to copy via current mirror in current field.

Figure 10 provides use one current-mirror structure to produce a demonstrative circuit of the multiple copies for a current signal of suing for peace further.Herein, these outputs (or output of these folders) of this symbol likelihood generator adopt current forms instead of voltage form.

For purposes of illustration, differential pair implements a folding amplifier (such as, the folders 820B of Fig. 8) with the combination of input IN1P, IN1M and IN2P, IN2M.IOUTP and IOUTM is the output current of this folding amplifier.IOUTPC1, IOUTPC2 and IOUTMC1, IOUTMC2 form the two copies of this IOUTP and IOUTM signal of deriving via " current reflection " operation respectively.Figure 11 A shows in order to produce two current signal I ₁and I ₂summation I ₃a demonstrative circuit.Figure 11 B illustrates the use of Kirchhoff s current law in the summation of this electric current.

4 extend and application

In this description, although this circuit example is mainly implemented to be illustrated in the background of the soft demapper function operated PAM modulation signal at use analog circuit, but this general technology easily can be applicable to other application many, and some of them are applied in and are hereafter described.

4.1 analog domains are to the mapping in probability territory

One application relates to analog domain to the signal in probability territory and videos, and such as analog signal (such as, electric current and voltage) is to the reflection of the probability of symbol and/or digit order number.These reflections, such as, can be when this analog signal comprises significant noise component(s) useful, make the probability measuring value of an estimated value that valuable information can be provided to do follow-up signal process and/or error correction.

Conclude according to this communication example, any observational variable (image pixel intensities, genome base equity) can be used as a reflector and any noise in this observation can be used as a channel.Any observation parameter (even a real number) can be considered the state be in one group of discrete state.In the example of a genome base-pair, there are 4 possible state G, C, T, A.One single-electronic transistor measures the configuration that can measure a neighbouring DNA transcriptase, but this will be a noise measurement.Therefore Received signal strength is by the discrete voltage levels in the discrete voltage levels of 4 for adding noise.In another example, we may want, and observation is known has a limited dynamic range and we wish the real number value variable known under a specified resolution.This variable can be considered as the state in possible a large amount of discrete states by us.Noise in measurement will again as a channel, and demapper is in the probability of each in this possible state or likelihood by producing this variable.

4.2 QAM and N tie up stellar map

One second application relates to and acts on QAM (quadrature amplitude modulation) modulation signal, or more generally, acts on the execution mode of the soft demapper function tieing up the signal that stellar map is modulated based on a N.

In general, QAM presentation mode uses two mathematically orthogonal signals (such as sinusoidal and cosine) to produce multiple modulation signal to carry out data transmission.In the case, transmission signal can be expressed as I (t) sin (ω t)+Q (t) cos (ω t), and wherein I (t) and Q (t) is called the quadrature component of this signal.

Figure 12 shows the QAM stellar map using Gray code.In this two-dimentional stellar map, adjacent symbols difference is each other no more than one in any position.For purposes of illustration, a qam symbol S _iQtransmission can be considered the transmission of two PAM symbol I and Q in two independent channels.Each PAM symbol has videos into two position (such as, b ₀and b ₁) four level (such as, I ₀, I ₁, I ₂and I ₃).The qam symbol produced has videos into 4 position (b ₀, b ₁, b ₂and b ₃) 16 level (S ₀₀to S ₃₃).

Assuming that interchannel noise, such as additive white Gaussian noise, affects this I and the Q component of this signal in an incoherent mode, or in other words, the transmission of I and Q is independently.Each symbol S in this QAM stellar map _iQthe symbol likelihood product that then can be used as the symbol likelihood of symbol I and Q corresponding in PAM stellar map obtain, such as P (S _1,3)=P (I ₁) P (Q ₃).Therefore, the independence that the calculating of the symbol likelihood of QAM stellar map can be reduced to for the respectively symbol likelihood of the PAM stellar map of this I and Q component calculates (the symbol likelihood generator 212 such as using Fig. 2).

Figure 13 display is configured to produce independent of the input ν transmitted in independent channel _iand ν _qthe example of a demapper circuit of symbol likelihood.

Herein, once obtain this symbol likelihood, position probability just with the limit comprising the set of the symbol likelihood of the corresponding value of selected position of this QAM stellar map and can calculate.In the ordinary course of things, position b _ithere is the probability of value " 1 ", such as, can obtain as follows:

$P(b_i=1)=\underset{k\∈K}{\underset{j\∈J}{\mathrm{\Σ}}}P\left(I_j\right)\·P\left(Q_k\right)---\left(2\right)$

Wherein J and K comprises a b from QAM stellar map _ithe set of all symbols of value " 1 ", and P (I _j) and P (Q _k) be symbol I _jand Q _krespective symbol likelihood.

In some examples, stellar map is formed with the calculating of reduced equation (2), such as, to avoid or to simplify direct multiplication with an ad hoc fashion.Please note, if equation (2) sum comprises the item of the product comprised from the special symbol of a PAM stellar map and all symbols of another PAM stellar map, then by the simple regular restrictive condition of application (namely all possible states of a stochastic variable probability and be 1) and distributive law, equation (2) can be rewritten as follows:

In this example of this Gray code stellar map of Figure 12, position b ₀for the probability of " 0 " is thus by calculating the summation of all stellar map symbols likelihood be included in gray area 1210, or in other words by calculating I ₀and I ₁the summation of area obtain.Therefore, b ₀for the probability of " 1 " is by by I ₂and I ₃area add up calculate.

Because Gray code is not one uniquely to represent, represent so Gray code QAM stellar map exists other.In addition, other coding techniquess also can be used for developing this stellar map.Without loss of generality, in order to marginalize, the position probability extended to for any multidimensional stellar map calculates by selecting suitable symbol likelihood set equation (2) and (3).In general, symbol likelihood calculating and position probability are calculated separate and allow this circuit flexibly to be configured again to use together with various stellar map or algorithm of videoing.

4.3 data storing application

3rd application relates to data storing application, such as, the stocking system (such as many level flash memory) that demapper and various memory device and the information of such as DRAM (Dynamic Random Access Memory) (DRAM), flash memory (comprising embedded flash memory) store as the variable with two or more quantization levels is wherein combined.

In some embodiments, this demapper can be arranged in the various positions on a memory chip, includes but not limited to be directly adjacent to an internal storage location, at the end points place of an internal memory row or column, or other positions of read path along this memory device.In such applications, this demapper can be used for the soft de-mapping/soft decoding being stored in the data in internal storage location, such as, to improve the bit error rate of this internal memory.For internal memory, this demapper implementing soft decoding method based on current sense amplifier also can help to improve reading speed relative to the internal memory of other types.

4.4 analog domains are to the mapping of rigid decision-making

4th application relates to the mapping of analog domain to rigid decision-making, such as can in order to soft input to be converted to the transducer of rigid output.Such as, a demapper can implement a quick flashing A/D, and each input range is videoed into a decision-making district respective in one group of discrete decision-making district by it, and wherein each decision-making district corresponds to a quantization level (such as, representing a rigid symbol or a rigid position).This demapper such as can use its transfer function to adopt a combination of the configurable tangent bend function generator of more vairable of nearly rectangle to configure.In some examples, the transfer function of transducer can adopt the form of a linear ladder.In some examples, the output of transducer can feed-in one digital encoder, and this input can be converted to a binary value by it.

5 other examples

The various alternate embodiments of this system and method are possible.

Such as, as discussed previously, the de-mapping device 210 of Fig. 2 implements the secondary simulation process of the likelihood of Continuous plus symbol and position probability.This mode allows to design the reconfigurable hardware that can be applicable to various stellar map or reflection algorithm, such as, by reusing a large amount of circuit unit (such as, producing the folders of various convex shaped or S shape transfer function).In some other examples, one substitutes de-mapping device can through building with the direct calculating of execute bit probability in one-level.Such as, when a specific stellar map is selected, its structure can be utilized the alternative configuration of this folders circuit 820A-820D determining Fig. 8, makes each folders use one group of suitable tangent bend function generator of more vairable to form the output signal directly corresponding to position probability (and not needing to produce symbol likelihood).This first-order can help this circuit to improve the computational efficiency of a target application, although be lower for cost with the flexibility of hardware configuration.One example of one single-stage demapper shows in this calcspar of Figure 14.Another example of one single-stage demapper shows in this calcspar of Figure 25.Other substitutes are also possible.

In certain embodiments, this de-mapping device can be configured to transmission signal is demodulated into symbol likelihood (and not necessarily producing soft position).This symbol likelihood then can be supplied to a subsequent conditioning circuit in order to further process.

In some applications, in circuit design, the differential signal being better than traditional single-ended signal is adopted may to be useful.The advantage of differential signal can comprise, and such as, has a comparatively great dynamic range and have higher noise vulnerability to jamming for circuit operation.In some examples, position probability can also difference form produce, such as:

DP(b ₀)＝P(b ₀＝1)-P(b ₀＝0) (4a)

DP(b ₁)＝P(b ₁＝1)-P(b ₁＝0) (4b)

DP(b ₂)＝P(b ₂＝1)-P(b ₂＝0) (4c)

DP(b ₃)＝P(b ₃＝1)-P(b ₃＝0) (4d)

Wherein DP (b _i) represent the difference probability of i-th position.

It is to be noted, in this circuit structure of Fig. 8, the characteristic that the symbol likelihood of each folders 820A-820D exports independently controls by the corresponding parameter adjusting input change-over circuit 810.Such as, parameter a ₁determine symbol grayscale curve P (S ₀) probability and parameter b ₁determine this slope of a curve.In some applications, the interval between the representative input value of distinct symbols grayscale curve may be not necessarily uniform.Each slope of a curve (or acutance) also may be different.This can be useful for some application, and such as, wherein distinct symbols may along with different noise characteristic by the application transmitted.Such as, end points symbol S ₃can use high voltage signal to transmit (being sometimes even in nearly saturation level), itself may carrying one comparatively very noisy component in some systems.The configurable degree of individual symbol grayscale curve enables a demapper customize according to for the particular channel of transfer of data or system.In addition, for internal memory or data storing application, the demapper operationally retrieving internal memory can come tuning particularly based on the noise/error characteristic of an internal storage location, a memory cell array or respective symbol.

In some applications, a noise estimator can combine with this demapper circuit of Fig. 8 the map feature implemented to control this de-mapping device based on noise estimation value.Such as, this noise estimator can analyze the characteristic of input signal ν (such as, the ratio of signal and noise) determine the shape (such as, acutance and interval) expected of this symbol grayscale curve and therefore calculate corresponding a and the b parameter that need be supplied to this demapper circuit.In addition, this noise estimator can implement a feedback mechanism, can come tuning based on the performance (such as, bit error rate) of a follow-up decoder circuit by this shape of this symbol grayscale curve of this feedback mechanism.

6 log-likelihood ratios (LLR) de-mapping device

In this explanation above-mentioned, the signal transmitted in channel uses and represents that the electric current of linear probability and/or voltage process in this demapper of an analog form.At some in other application, the signal received is selectively treated to be represented with the probability produced in log-domain, such as, and logarithm stratagem ensuring success ratio or log-likelihood ratio (LLR).The log-likelihood ratio (LLR) of one variable x may be defined as or be selectively defined as for purposes of discussion, LLR will be used _xfirst definition.

In one embodiment, the LLR that soft demapper circuit is configured to successive value input signal to convert to symbol or position probability represents.

With reference to Figure 15, show one group of 8 Gaussian Profile of the voltage signal from PAM-8 communication channel.The corresponding LLR curve calculated of Figure 16 show needle contraposition 0 (curve 160), position 1 (curve 162) and position 2 (curve 164).From a circuit power and area visual angle, due at least two reasons, use analog circuit accurately to copy these LLR curves and may be difficulty and costliness.First, the discontinuity implementing the sharp transition (1: 166 place such as, on curve 160) between such as positive slope and negative slope is difficult.Secondly, the afterbody of the LLR curve that this calculates extends to the infinity (such as, in region 168) exceeded outside the boundary value converting this group signal level that a probability represents to.

The core of this LLR signal approximate is generally enough to relevant probability information to pass to a decoder or demodulator in order to efficiently and accurately to calculate, instead of the Precise Representation (such as, by the segmentation superposition of linear segment) of this LLR signal of generation Figure 16.By carrying out this LLR curve of approximate Figure 16 and high value and the lower value by the limit being forced at this curve to have the level and smooth curve changed, can obtain comprise insufficient information with to realize in follow-up decoding or demodulation step the curve of approximation of expectation performance.

With reference to Figure 17 A-17C, approximate LLR curve 170,172 and 174 corresponds respectively to accurate LLR curve 160,162 and 164.Transformation between positive slope and negative slope is level and smooth for this curve of approximation and this curve is demarcated not extend to infinity.With reference to Figure 18, show this three approximate LLR curve 170,172 and 174.The limit of each curve is that staggered the make upper bound of curve 172 (b=1) and the value of lower bound is greater than the upper bound of curve 170 (b=0) and the value of lower bound slightly.Similarly, the upper bound of curve 174 (b=2) and the value of lower bound are greater than the upper bound of curve 172 (b=1) and the value of lower bound slightly.In general, for the curve of arbitrary number, these limit are that the staggered limit of each continuous position that makes is greater than the limit of first anteposition gradually.

In order to implement approximate LLR curve 170,172 and 174 in analog circuit, it is useful for recognizing that each curve of approximation can be built by the mirror image of tangent bend function shape and tangent bend function shape.For example, referring to Figure 19, curve 170 (b=0) is decomposed into the section 190,192,194 and 196 of four tangent bend function shape.

With reference to Figure 20, respectively this tangent bend function section can be implemented in the differential pair circuit 250 comprising the tangent bend function generator of more vairable 256.Differential pair circuit 250 also comprises the gain component such as each tangent bend function section being suitably positioned the adder 252 in x-axis and adjusting the side-play amount of this tangent bend function section and a variable gain amplifier 254 of slope.Such as, the LLR slope of curve with the change of different additive white Gaussian noise (AWGN) characteristic suitably sets by one of variable gain amplifier 254 and processes.With regard to curve 170,4 of circuit 250 illustrate in order to produce two double curved shape (corresponding to section 192 and 196) and two mirror images (corresponding to section 190 and 194).Curve 172 (b=1) can utilize two of circuit 250 illustrations to implement; Curve 174 (b=2) can utilize circuit 250 illustration to implement.21A and 21B figure is respectively provided in the example of an adder circuit and the variable gain circuit used in circuit 250.Other Circnit Layouts are also possible.

With reference to Figure 26, in another embodiment, differential pair circuit 2650 is in order to implement each tangent bend function section.Differential pair circuit 2650 comprises an adder 2652, the tangent bend function generator of more vairable 2656 and a variable gain amplifier 2654.This adder respectively in 21A and 21B figure and variable gain circuit can use in differential pair circuit 2650.

In this tangent bend function generator of more vairable above-mentioned, this electric current exports the current signal that carrying only has single polarity.As shown in 21A and 21B figure, output connecting pin is the current sink that electric current flows into; When no current source, current flowing is to outside this output connecting pin.In order to the current waveform produced by these tangent bend function generators of more vairable is added to produce a single LLR crest (hump), curve in such as 17B figure 172, two curved sections (mirror image of a pair of function of flexure and a pair of function of flexure) can be added without adjustment.

In order to produce the extra bending of multiple tangent bend sections of such as curve 170 in this curve, extra curved section must be added into, and it changes the DC location of final curve of output.But, there is multiple bending curve in order to implement higher-order PAM code signal or in order to implement in one dimension more than two positions.Two kinds of modes can be taked to regulate this DC side-play amount.In mode, DC can be added and adjust electric current.In second method, current mirror in order to produce the current waveform of opposite polarity, thus can be avoided this DC in outputing signal to offset, but introduces extra complexity in the circuit.

Only the common mode electrical level of output current signal will have this DC component.This differential signal obtained by above-mentioned which has practically negligible acceptable common mode.For the common mode of the indivedual single-ended side of this signal, preferably this DC component can not be increased to and can accept in order to form the extraneous level of interface with other circuit (such as follow-up decoder).

In other embodiments, the symbol likelihood generator 800 of Fig. 8 can be configured in being exercisable under number state, such as, by by each folders with convex shaped or S pictograph likelihood transfer function to be converted in log-domain expect that the additional circuit components of form couples.Such as, a variable gain amplifier can add to amplify this output in each probability output so that approximate corresponding LLR slope of a curve.In addition, suitably staggered limiter can add this value to limit this output at character boundary place.But these current components may be expensive.In an alternative em bodiment, the symbol likelihood generator 800 of Fig. 8 can be configured in being exercisable under number state, such as, by only setting parameter a and b with transfer function needed for optimal approximation.

It is to be noted, this communication system described herein is understood in an extensive background, and is not limited to telecommunication system or communication network.In addition, these method and system are generally applicable to the region that wherein statistical inference technology is suitable for from the extracting data information damaged by various noise source and/or distortion sources.

Be understood that this explanation aforementioned is intended to illustrate and do not limit the scope of the present invention defined by the scope of claims.Other embodiments are in the scope of following claims.

Accompanying drawing explanation

Fig. 1 is the calcspar of communication system.

Fig. 2 is the calcspar of the exemplary simulation execution mode of the receiver of this communication system for Fig. 1.

Fig. 3 A and 3B is respectively as the curve chart of symbol likelihood of function of input signal of two channels with different noise characteristic.

Fig. 4 illustrates that use a pair tangent bend function generator of more vairable is to produce the mode of convex shaped transfer function.

Fig. 5 A is to provide the calcspar of the circuit implementation of this transfer function of Fig. 4.

Fig. 5 B is the circuit diagram of the example of the folders of Fig. 5 A.

Fig. 6 illustrates the mode using the tangent bend function generator of more vairable to produce displacement tangent bend transfer function.

Fig. 7 A is to provide the calcspar of a circuit implementation of this transfer function of Fig. 6.

Fig. 7 B is the circuit diagram of an example of the folders of Fig. 7 A.

Fig. 8 is the calcspar of an embodiment of a symbol likelihood generator.

Fig. 9 is the calcspar of an embodiment of the probability generator used together with this symbol likelihood generator of Fig. 8.

Figure 10 is the circuit diagram of the multiple copies producing a current signal.

Figure 11 A and 11B illustrates an electric current summing circuit.

Figure 12 illustrates a quadrature amplitude modulation (QAM) stellar map.

Figure 13 illustrates another embodiment of the symbol likelihood generator used together with this QAM stellar map of Figure 12.

Figure 14 illustrates an embodiment of one-level de-mapping device.

Figure 15 is the curve chart of the voltage's distribiuting of 8 voltage signals from PAM-8 communication channel.

Figure 16 is the curve chart of log-likelihood ratio (LLR) curve of this voltage signal corresponding to Figure 15.

Figure 17 A-17C is this accurate LLR curve of Figure 16 and the curve chart of this corresponding approximate LLR curve.

Figure 18 is the curve chart of this approximate LLR curve of Figure 17 A-17C.

Figure 19 is the curve chart of this approximate LLR curve corresponding to position 0.

Figure 20 is the calcspar of the circuit implementation of this approximate LLR curve producing Figure 19.

21A and 21B figure is the circuit diagram of an exemplary tangent bend shifter circuit and the exemplary variable gain circuit used in this circuit implementation of Figure 20 respectively.

Figure 22 A illustrates that use a pair tangent bend function generator of more vairable produces one second embodiment of a convex shaped transfer function.

Figure 22 B is to provide the calcspar of a circuit implementation of this transfer function of Figure 22 A.

Figure 23 A is one second embodiment using the tangent bend function generator of more vairable to produce displacement tangent bend transfer function.

Figure 23 B is to provide the calcspar of a circuit implementation of this transfer function of Figure 23 A.

Figure 24 is the calcspar of one second embodiment of a symbol likelihood generator.

Figure 25 illustrates one second embodiment of one-level de-mapping device.

Figure 26 is the calcspar of one second embodiment of a circuit of this approximate LLR curve producing Figure 19.

Claims (30)

1. a circuit, it comprises:

Signal processing circuit, it is in order to reception input and in order to produce one group of output, this input is in the input range comprising a class value, and each exports the measuring value associated representing this input and one or more typical value in described typical value, and this signal processing circuit comprises:

Multiple output, each output is in response to this input of this signal processing circuit, and each output comprises:

One or more tangent bend function generator of more vairable, each tangent bend function generator of more vairable produces the output of the tangent bend function of this input representing this output in response to the input of this output; And

A circuit exported during described output in order to combine this one or more tangent bend function generator of more vairable exports to this group forming this signal processing circuit; And

Input change-over circuit, it is coupled to described multiple output, this input configurablely in order to change this signal processing circuit of this input change-over circuit control from this input to this group export map feature.

2. circuit as claimed in claim 1, wherein this output of this tangent bend function generator of more vairable represents the second tangent bend function of this input arriving this signal processing circuit.

3. circuit as claimed in claim 2, this map feature wherein inputing to the output of this group from this is associated with the characteristic of this second tangent bend function.

4. circuit as claimed in claim 3, wherein this characteristic of this second tangent bend function comprises the position of the transition point of this tangent bend function.

5. circuit as claimed in claim 3, wherein this characteristic of this second tangent bend function comprises the rate of change at the transition point place of this tangent bend function.

6. circuit as claimed in claim 1, wherein at least one output comprises two tangent bend function generators of more vairable.

7. circuit as claimed in claim 1, wherein each output also comprises: output conversion circuit, and it controls to input to from this second map feature that this group exports in order to this output of this one or more tangent bend function generator of more vairable corresponding to photograph configurablely.

8. circuit as claimed in claim 1, wherein each tangent bend function generator of more vairable comprises differential pair circuit.

9. circuit as claimed in claim 8, wherein this differential pair circuit comprises pair of transistor, and each transistor is configured this input receiving this tangent bend function generator of more vairable with the corresponding polarity in bipolarity.

10. circuit as claimed in claim 8, wherein this differential pair circuit comprises a pair mosfet transistor being configured to operate under an overcritical Value Operations state.

11. circuit as claimed in claim 8, wherein this differential pair circuit comprises a pair bipolar junction transistor.

12. circuit as claimed in claim 1, this input wherein arriving this signal processing circuit comprises and representing in communication channel with the analog signal of the numeric string of sign format transmission.

13. circuit as claimed in claim 12, wherein each typical value of this input range is associated to the corresponding symbol in discrete one group of possible symbol.

14. circuit as claimed in claim 13, wherein each output of this signal processing circuit represents that this transmission symbol corresponds to the corresponding probability of the symbol in described possibility symbol.

15. circuit as claimed in claim 13, each of wherein this signal processing circuit exports and represents that one of the numeric string corresponding numeral transmitted has the corresponding probability of the probable value in one group of discrete probable value.

16. circuit as claimed in claim 15, wherein transmitted numeric string is binary string.

17. circuit as claimed in claim 13, wherein each output of this signal processing circuit comprises the expression of transmitted position in log-domain.

18. circuit as claimed in claim 12, wherein this analog signal comprises at least partly by the determined noise component(s) of this communication channel.

19. circuit as claimed in claim 18, wherein this analog signal comprises Gaussian noise component.

20. circuit as claimed in claim 1, this input wherein arriving this signal processing circuit is implemented with the form of differential signal.

21. circuit as claimed in claim 1, each input wherein arriving described output is implemented with the form of differential signal.

22. 1 kinds of signal processing circuits, it is in order to reception input and in order to produce one group of output, this input is in the input range comprising a class value, and each exports the measuring value associated representing this input and one or more typical value in described typical value, and this signal processing circuit comprises:

Multiple part, every part produces the corresponding output that this group exports, and every part comprises:

One or more inputs change-over circuit, and this one or more input change-over circuit this input configurablely in order to change this signal processing circuit controls to input to the map feature of this corresponding output from this;

One or more tangent bend function generator of more vairable, each tangent bend function generator of more vairable is coupled to corresponding input change-over circuit and is configured to receive the input after conversion from this corresponding input change-over circuit and produce the tangent bend function output of the tangent bend function representing this input; And

In order to combination from the described tangent bend function of the described tangent bend function generator of more vairable export this group forming this signal processing circuit export in a circuit exported.

23. signal processing circuits as claimed in claim 22, each output wherein in the output of this group comprises the expression of this input in log-domain.

24. signal processing circuits as claimed in claim 23, each wherein in the output of this group exports the log-likelihood ratio representing this input.

25. signal processing circuits as claimed in claim 22, wherein every part also comprises one or more output conversion circuit, and each output conversion circuit is coupled to the corresponding tangent bend function generator of more vairable and exports in order to this tangent bend function that photograph is corresponding configurablely and controls to input to the second map feature of this corresponding output from this.

26. 1 kinds in order to reception input and in order to produce one group of method exported, this input is in the input range comprising a class value, and each exports the measuring value associated representing this input and one or more typical value in described typical value, the method includes the steps of:

Use input change-over circuit change this input with control from this input to this group export map feature;

The following step is performed in each output of multiple output:

At least one the tangent bend function producing the tangent bend function of the input after representing conversion exports, and each tangent bend function exports and produced by the corresponding tangent bend function generator of more vairable; And

Use combinational circuit to combine described tangent bend function to export with the output formed in the output of this group.

27. methods as claimed in claim 26, wherein this output of this tangent bend function generator of more vairable represents the second tangent bend function of this input.

28. methods as claimed in claim 27, this map feature wherein inputing to the output of this group from this is associated with the characteristic of this second tangent bend function.

29. methods as claimed in claim 28, wherein this characteristic of this second tangent bend function comprise the rate of change at the position of the transition point of this tangent bend function and the transition point place of this tangent bend function one of at least.

30. methods as claimed in claim 26, it also comprises, in each output of described multiple output, use corresponding output conversion circuit to change each tangent bend function export with control from this input to this group export the second map feature.